if_stf.c revision 87473
1/*	$FreeBSD: head/sys/net/if_stf.c 87473 2001-12-07 01:32:40Z arr $	*/
2/*	$KAME: if_stf.c,v 1.62 2001/06/07 22:32:16 itojun Exp $	*/
3
4/*
5 * Copyright (C) 2000 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33/*
34 * 6to4 interface, based on RFC3056.
35 *
36 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
37 * There is no address mapping defined from IPv6 multicast address to IPv4
38 * address.  Therefore, we do not have IFF_MULTICAST on the interface.
39 *
40 * Due to the lack of address mapping for link-local addresses, we cannot
41 * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
42 * packets to link-local multicast addresses (ff02::x).
43 *
44 * Here are interesting symptoms due to the lack of link-local address:
45 *
46 * Unicast routing exchange:
47 * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
48 *   and link-local addresses as nexthop.
49 * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
50 *   assigned to the link, and makes use of them.  Also, HELLO packets use
51 *   link-local multicast addresses (ff02::5 and ff02::6).
52 * - BGP4+: Maybe.  You can only use global address as nexthop, and global
53 *   address as TCP endpoint address.
54 *
55 * Multicast routing protocols:
56 * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
57 *   Adjacent PIM routers must be configured manually (is it really spec-wise
58 *   correct thing to do?).
59 *
60 * ICMPv6:
61 * - Redirects cannot be used due to the lack of link-local address.
62 *
63 * stf interface does not have, and will not need, a link-local address.
64 * It seems to have no real benefit and does not help the above symptoms much.
65 * Even if we assign link-locals to interface, we cannot really
66 * use link-local unicast/multicast on top of 6to4 cloud (since there's no
67 * encapsulation defined for link-local address), and the above analysis does
68 * not change.  RFC3056 does not mandate the assignment of link-local address
69 * either.
70 *
71 * 6to4 interface has security issues.  Refer to
72 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
73 * for details.  The code tries to filter out some of malicious packets.
74 * Note that there is no way to be 100% secure.
75 */
76
77#include "opt_inet.h"
78#include "opt_inet6.h"
79
80#include <sys/param.h>
81#include <sys/systm.h>
82#include <sys/socket.h>
83#include <sys/sockio.h>
84#include <sys/mbuf.h>
85#include <sys/errno.h>
86#include <sys/kernel.h>
87#include <sys/protosw.h>
88#include <sys/queue.h>
89#include <machine/bus.h>	/* XXX: Shouldn't really be required! */
90#include <sys/rman.h>
91#include <machine/cpu.h>
92
93#include <sys/malloc.h>
94
95#include <net/if.h>
96#include <net/route.h>
97#include <net/netisr.h>
98#include <net/if_types.h>
99#include <net/if_stf.h>
100
101#include <netinet/in.h>
102#include <netinet/in_systm.h>
103#include <netinet/ip.h>
104#include <netinet/ip_var.h>
105#include <netinet/in_var.h>
106
107#include <netinet/ip6.h>
108#include <netinet6/ip6_var.h>
109#include <netinet6/in6_var.h>
110#include <netinet/ip_ecn.h>
111
112#include <netinet/ip_encap.h>
113
114#include <machine/stdarg.h>
115
116#include <net/net_osdep.h>
117
118#include <net/bpf.h>
119
120#define STFNAME		"stf"
121#define STF_MAXUNIT	0	/* only one is currently allowed */
122
123#define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
124#define GET_V4(x)	((struct in_addr *)(&(x)->s6_addr16[1]))
125
126struct stf_softc {
127	struct ifnet	sc_if;	   /* common area */
128	union {
129		struct route  __sc_ro4;
130		struct route_in6 __sc_ro6; /* just for safety */
131	} __sc_ro46;
132#define sc_ro	__sc_ro46.__sc_ro4
133	const struct encaptab *encap_cookie;
134	struct resource *r_unit;	/* resource allocated for this unit */
135	LIST_ENTRY(stf_softc) sc_list;	/* all stf's are linked */
136};
137
138LIST_HEAD(, stf_softc) stf_softc_list;
139
140static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
141static struct rman stfunits[1];
142static int ip_stf_ttl = 40;
143
144extern  struct domain inetdomain;
145struct protosw in_stf_protosw =
146{ SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
147  in_stf_input, rip_output,	0,		rip_ctloutput,
148  0,
149  0,            0,              0,              0,
150  &rip_usrreqs
151};
152
153static int stfmodevent __P((module_t, int, void *));
154static int stf_encapcheck __P((const struct mbuf *, int, int, void *));
155static struct in6_ifaddr *stf_getsrcifa6 __P((struct ifnet *));
156static int stf_output __P((struct ifnet *, struct mbuf *, struct sockaddr *,
157	struct rtentry *));
158static int stf_checkaddr4 __P((struct stf_softc *, struct in_addr *,
159	struct ifnet *));
160static int stf_checkaddr6 __P((struct stf_softc *, struct in6_addr *,
161	struct ifnet *));
162static void stf_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *));
163static int stf_ioctl __P((struct ifnet *, u_long, caddr_t));
164
165int	stf_clone_create __P((struct if_clone *, int *));
166void	stf_clone_destroy __P((struct ifnet *));
167
168struct if_clone stf_cloner =
169    IF_CLONE_INITIALIZER(STFNAME, stf_clone_create, stf_clone_destroy);
170
171int
172stf_clone_create(ifc, unit)
173	struct if_clone *ifc;
174	int *unit;
175{
176	struct resource *r;
177	struct stf_softc *sc;
178
179	if (*unit > STF_MAXUNIT)
180		return (ENXIO);
181
182	if (*unit < 0) {
183		 r = rman_reserve_resource(stfunits, 0, STF_MAXUNIT, 1,
184		     RF_ALLOCATED | RF_ACTIVE, NULL);
185		 if (r == NULL)
186			return (ENOSPC);
187		 *unit = rman_get_start(r);
188	} else {
189		r = rman_reserve_resource(stfunits, *unit, *unit, 1,
190		    RF_ALLOCATED | RF_ACTIVE, NULL);
191		if (r == NULL)
192			 return (EEXIST);
193	}
194
195	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
196	sc->sc_if.if_name = STFNAME;
197	sc->sc_if.if_unit = *unit;
198	sc->r_unit = r;
199
200	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
201	    stf_encapcheck, &in_stf_protosw, sc);
202	if (sc->encap_cookie == NULL) {
203		printf("%s: attach failed\n", if_name(&sc->sc_if));
204		free(sc, M_STF);
205		return (ENOMEM);
206	}
207
208	sc->sc_if.if_mtu    = IPV6_MMTU;
209	sc->sc_if.if_ioctl  = stf_ioctl;
210	sc->sc_if.if_output = stf_output;
211	sc->sc_if.if_type   = IFT_STF;
212	sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
213	if_attach(&sc->sc_if);
214	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
215	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
216	return (0);
217}
218
219void
220stf_clone_destroy(ifp)
221	struct ifnet *ifp;
222{
223	int err;
224	struct stf_softc *sc = (void *) ifp;
225
226	LIST_REMOVE(sc, sc_list);
227	err = encap_detach(sc->encap_cookie);
228	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
229	bpfdetach(ifp);
230	if_detach(ifp);
231
232	err = rman_release_resource(sc->r_unit);
233	KASSERT(err == 0, ("Unexpected error freeing resource"));
234
235	free(sc, M_STF);
236}
237
238static int
239stfmodevent(mod, type, data)
240	module_t mod;
241	int type;
242	void *data;
243{
244	int err;
245
246	switch (type) {
247	case MOD_LOAD:
248		stfunits->rm_type = RMAN_ARRAY;
249		stfunits->rm_descr = "configurable if_stf units";
250		err = rman_init(stfunits);
251		if (err != 0)
252			return (err);
253		err = rman_manage_region(stfunits, 0, STF_MAXUNIT);
254		if (err != 0) {
255			printf("%s: stfunits: rman_manage_region: Failed %d\n",
256			    STFNAME, err);
257			rman_fini(stfunits);
258			return (err);
259		}
260		LIST_INIT(&stf_softc_list);
261		if_clone_attach(&stf_cloner);
262
263		break;
264	case MOD_UNLOAD:
265		if_clone_detach(&stf_cloner);
266
267		while (!LIST_EMPTY(&stf_softc_list))
268			stf_clone_destroy(&LIST_FIRST(&stf_softc_list)->sc_if);
269
270		err = rman_fini(stfunits);
271		KASSERT(err == 0, ("Unexpected error freeing resource"));
272
273		break;
274	}
275
276	return (0);
277}
278
279static moduledata_t stf_mod = {
280	"if_stf",
281	stfmodevent,
282	0
283};
284
285DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
286
287static int
288stf_encapcheck(m, off, proto, arg)
289	const struct mbuf *m;
290	int off;
291	int proto;
292	void *arg;
293{
294	struct ip ip;
295	struct in6_ifaddr *ia6;
296	struct stf_softc *sc;
297	struct in_addr a, b;
298
299	sc = (struct stf_softc *)arg;
300	if (sc == NULL)
301		return 0;
302
303	if ((sc->sc_if.if_flags & IFF_UP) == 0)
304		return 0;
305
306	/* IFF_LINK0 means "no decapsulation" */
307	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
308		return 0;
309
310	if (proto != IPPROTO_IPV6)
311		return 0;
312
313	/* LINTED const cast */
314	m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
315
316	if (ip.ip_v != 4)
317		return 0;
318
319	ia6 = stf_getsrcifa6(&sc->sc_if);
320	if (ia6 == NULL)
321		return 0;
322
323	/*
324	 * check if IPv4 dst matches the IPv4 address derived from the
325	 * local 6to4 address.
326	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
327	 */
328	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
329	    sizeof(ip.ip_dst)) != 0)
330		return 0;
331
332	/*
333	 * check if IPv4 src matches the IPv4 address derived from the
334	 * local 6to4 address masked by prefixmask.
335	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
336	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
337	 */
338	bzero(&a, sizeof(a));
339	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
340	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
341	b = ip.ip_src;
342	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
343	if (a.s_addr != b.s_addr)
344		return 0;
345
346	/* stf interface makes single side match only */
347	return 32;
348}
349
350static struct in6_ifaddr *
351stf_getsrcifa6(ifp)
352	struct ifnet *ifp;
353{
354	struct ifaddr *ia;
355	struct in_ifaddr *ia4;
356	struct sockaddr_in6 *sin6;
357	struct in_addr in;
358
359	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
360	     ia;
361	     ia = TAILQ_NEXT(ia, ifa_list))
362	{
363		if (ia->ifa_addr == NULL)
364			continue;
365		if (ia->ifa_addr->sa_family != AF_INET6)
366			continue;
367		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
368		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
369			continue;
370
371		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
372		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
373			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
374				break;
375		if (ia4 == NULL)
376			continue;
377
378		return (struct in6_ifaddr *)ia;
379	}
380
381	return NULL;
382}
383
384static int
385stf_output(ifp, m, dst, rt)
386	struct ifnet *ifp;
387	struct mbuf *m;
388	struct sockaddr *dst;
389	struct rtentry *rt;
390{
391	struct stf_softc *sc;
392	struct sockaddr_in6 *dst6;
393	struct in_addr *in4;
394	struct sockaddr_in *dst4;
395	u_int8_t tos;
396	struct ip *ip;
397	struct ip6_hdr *ip6;
398	struct in6_ifaddr *ia6;
399
400	sc = (struct stf_softc*)ifp;
401	dst6 = (struct sockaddr_in6 *)dst;
402
403	/* just in case */
404	if ((ifp->if_flags & IFF_UP) == 0) {
405		m_freem(m);
406		return ENETDOWN;
407	}
408
409	/*
410	 * If we don't have an ip4 address that match my inner ip6 address,
411	 * we shouldn't generate output.  Without this check, we'll end up
412	 * using wrong IPv4 source.
413	 */
414	ia6 = stf_getsrcifa6(ifp);
415	if (ia6 == NULL) {
416		m_freem(m);
417		return ENETDOWN;
418	}
419
420	if (m->m_len < sizeof(*ip6)) {
421		m = m_pullup(m, sizeof(*ip6));
422		if (!m)
423			return ENOBUFS;
424	}
425	ip6 = mtod(m, struct ip6_hdr *);
426	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
427
428	/*
429	 * Pickup the right outer dst addr from the list of candidates.
430	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
431	 */
432	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
433		in4 = GET_V4(&ip6->ip6_dst);
434	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
435		in4 = GET_V4(&dst6->sin6_addr);
436	else {
437		m_freem(m);
438		return ENETUNREACH;
439	}
440
441#if NBPFILTER > 0
442	if (ifp->if_bpf) {
443		/*
444		 * We need to prepend the address family as
445		 * a four byte field.  Cons up a dummy header
446		 * to pacify bpf.  This is safe because bpf
447		 * will only read from the mbuf (i.e., it won't
448		 * try to free it or keep a pointer a to it).
449		 */
450		struct mbuf m0;
451		u_int32_t af = AF_INET6;
452
453		m0.m_next = m;
454		m0.m_len = 4;
455		m0.m_data = (char *)&af;
456
457#ifdef HAVE_OLD_BPF
458		bpf_mtap(ifp, &m0);
459#else
460		bpf_mtap(ifp->if_bpf, &m0);
461#endif
462	}
463#endif /*NBPFILTER > 0*/
464
465	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
466	if (m && m->m_len < sizeof(struct ip))
467		m = m_pullup(m, sizeof(struct ip));
468	if (m == NULL)
469		return ENOBUFS;
470	ip = mtod(m, struct ip *);
471
472	bzero(ip, sizeof(*ip));
473
474	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
475	    &ip->ip_src, sizeof(ip->ip_src));
476	bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
477	ip->ip_p = IPPROTO_IPV6;
478	ip->ip_ttl = ip_stf_ttl;
479	ip->ip_len = m->m_pkthdr.len;	/*host order*/
480	if (ifp->if_flags & IFF_LINK1)
481		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
482	else
483		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
484
485	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
486	if (dst4->sin_family != AF_INET ||
487	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
488		/* cache route doesn't match */
489		dst4->sin_family = AF_INET;
490		dst4->sin_len = sizeof(struct sockaddr_in);
491		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
492		if (sc->sc_ro.ro_rt) {
493			RTFREE(sc->sc_ro.ro_rt);
494			sc->sc_ro.ro_rt = NULL;
495		}
496	}
497
498	if (sc->sc_ro.ro_rt == NULL) {
499		rtalloc(&sc->sc_ro);
500		if (sc->sc_ro.ro_rt == NULL) {
501			m_freem(m);
502			return ENETUNREACH;
503		}
504	}
505
506	return ip_output(m, NULL, &sc->sc_ro, 0, NULL);
507}
508
509static int
510stf_checkaddr4(sc, in, inifp)
511	struct stf_softc *sc;
512	struct in_addr *in;
513	struct ifnet *inifp;	/* incoming interface */
514{
515	struct in_ifaddr *ia4;
516
517	/*
518	 * reject packets with the following address:
519	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
520	 */
521	if (IN_MULTICAST(ntohl(in->s_addr)))
522		return -1;
523	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
524	case 0: case 127: case 255:
525		return -1;
526	}
527
528	/*
529	 * reject packets with broadcast
530	 */
531	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
532	     ia4;
533	     ia4 = TAILQ_NEXT(ia4, ia_link))
534	{
535		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
536			continue;
537		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
538			return -1;
539	}
540
541	/*
542	 * perform ingress filter
543	 */
544	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
545		struct sockaddr_in sin;
546		struct rtentry *rt;
547
548		bzero(&sin, sizeof(sin));
549		sin.sin_family = AF_INET;
550		sin.sin_len = sizeof(struct sockaddr_in);
551		sin.sin_addr = *in;
552		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
553		if (!rt || rt->rt_ifp != inifp) {
554#if 0
555			log(LOG_WARNING, "%s: packet from 0x%x dropped "
556			    "due to ingress filter\n", if_name(&sc->sc_if),
557			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
558#endif
559			if (rt)
560				rtfree(rt);
561			return -1;
562		}
563		rtfree(rt);
564	}
565
566	return 0;
567}
568
569static int
570stf_checkaddr6(sc, in6, inifp)
571	struct stf_softc *sc;
572	struct in6_addr *in6;
573	struct ifnet *inifp;	/* incoming interface */
574{
575	/*
576	 * check 6to4 addresses
577	 */
578	if (IN6_IS_ADDR_6TO4(in6))
579		return stf_checkaddr4(sc, GET_V4(in6), inifp);
580
581	/*
582	 * reject anything that look suspicious.  the test is implemented
583	 * in ip6_input too, but we check here as well to
584	 * (1) reject bad packets earlier, and
585	 * (2) to be safe against future ip6_input change.
586	 */
587	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
588		return -1;
589
590	return 0;
591}
592
593void
594in_stf_input(m, off)
595	struct mbuf *m;
596	int off;
597{
598	int proto;
599	struct stf_softc *sc;
600	struct ip *ip;
601	struct ip6_hdr *ip6;
602	u_int8_t otos, itos;
603	int len, isr;
604	struct ifqueue *ifq = NULL;
605	struct ifnet *ifp;
606
607	proto = mtod(m, struct ip *)->ip_p;
608
609	if (proto != IPPROTO_IPV6) {
610		m_freem(m);
611		return;
612	}
613
614	ip = mtod(m, struct ip *);
615
616	sc = (struct stf_softc *)encap_getarg(m);
617
618	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
619		m_freem(m);
620		return;
621	}
622
623	ifp = &sc->sc_if;
624
625	/*
626	 * perform sanity check against outer src/dst.
627	 * for source, perform ingress filter as well.
628	 */
629	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
630	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
631		m_freem(m);
632		return;
633	}
634
635	otos = ip->ip_tos;
636	m_adj(m, off);
637
638	if (m->m_len < sizeof(*ip6)) {
639		m = m_pullup(m, sizeof(*ip6));
640		if (!m)
641			return;
642	}
643	ip6 = mtod(m, struct ip6_hdr *);
644
645	/*
646	 * perform sanity check against inner src/dst.
647	 * for source, perform ingress filter as well.
648	 */
649	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
650	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
651		m_freem(m);
652		return;
653	}
654
655	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
656	if ((ifp->if_flags & IFF_LINK1) != 0)
657		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
658	else
659		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
660	ip6->ip6_flow &= ~htonl(0xff << 20);
661	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
662
663	m->m_pkthdr.rcvif = ifp;
664
665	if (ifp->if_bpf) {
666		/*
667		 * We need to prepend the address family as
668		 * a four byte field.  Cons up a dummy header
669		 * to pacify bpf.  This is safe because bpf
670		 * will only read from the mbuf (i.e., it won't
671		 * try to free it or keep a pointer a to it).
672		 */
673		struct mbuf m0;
674		u_int32_t af = AF_INET6;
675
676		m0.m_next = m;
677		m0.m_len = 4;
678		m0.m_data = (char *)&af;
679
680#ifdef HAVE_OLD_BPF
681		bpf_mtap(ifp, &m0);
682#else
683		bpf_mtap(ifp->if_bpf, &m0);
684#endif
685	}
686
687	/*
688	 * Put the packet to the network layer input queue according to the
689	 * specified address family.
690	 * See net/if_gif.c for possible issues with packet processing
691	 * reorder due to extra queueing.
692	 */
693	ifq = &ip6intrq;
694	isr = NETISR_IPV6;
695
696	len = m->m_pkthdr.len;
697	if (! IF_HANDOFF(ifq, m, NULL))
698		return;
699	schednetisr(isr);
700	ifp->if_ipackets++;
701	ifp->if_ibytes += len;
702}
703
704/* ARGSUSED */
705static void
706stf_rtrequest(cmd, rt, info)
707	int cmd;
708	struct rtentry *rt;
709	struct rt_addrinfo *info;
710{
711
712	if (rt)
713		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
714}
715
716static int
717stf_ioctl(ifp, cmd, data)
718	struct ifnet *ifp;
719	u_long cmd;
720	caddr_t data;
721{
722	struct ifaddr *ifa;
723	struct ifreq *ifr;
724	struct sockaddr_in6 *sin6;
725	int error;
726
727	error = 0;
728	switch (cmd) {
729	case SIOCSIFADDR:
730		ifa = (struct ifaddr *)data;
731		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
732			error = EAFNOSUPPORT;
733			break;
734		}
735		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
736		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
737			ifa->ifa_rtrequest = stf_rtrequest;
738			ifp->if_flags |= IFF_UP;
739		} else
740			error = EINVAL;
741		break;
742
743	case SIOCADDMULTI:
744	case SIOCDELMULTI:
745		ifr = (struct ifreq *)data;
746		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
747			;
748		else
749			error = EAFNOSUPPORT;
750		break;
751
752	default:
753		error = EINVAL;
754		break;
755	}
756
757	return error;
758}
759